CNFinBench
Models answer financial questions and carry out financial workflows. The benchmark includes tasks on expertise, autonomy, and integrity.
- Made by
- OpenCompass and collaborators
- Released
- December 2025
- Kind
- Finance
- Level
- Expert
- Scored
- Settled by events
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Risk
- Privacy, Specialized advice
- Subject
- Economics and business, Law
- Language
- Chinese
- Interaction
- Agentic
- Environment
- APIs and tools
- Items from
- Real work and use, Generated
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Agents, Law and finance, Content safety
- Strongest model tested
- Claude Opus 5.5 #5 on the capability index
- Last model added
- September 2026
- Scores refreshed
- November 20th, 2026
Doubao Seed 2.1 Pro leads with 83.7 among models released by September 2026 Doubao Seed 2.1 Pro leads with 83.7 among models released by September 2026
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6.1 Terra
- and 13 more
- 1
Doubao Seed 2.1 Pro LLM 83.7
- 2 Kimi K3 LLM 82.9
- 3
DeepSeek V4.1 Flash LLM 80.6
- 4 GPT-6.1 Sol LLM 80.6
- 5 Intern-S2-Preview-397B LLM 80.3
- 6
Gemini 3.1 Pro LLM 80.0
- 7 Claude Opus 5.5 LLM 79.1
- 8 GPT-5.6 Sol Agent 78.1
- 9 MTFin-Qwen3.8-V2 LLM 76.0
- 10
Qwen 3.8 Max LLM 74.9
- 11
MiniMax-M3 LLM 69.5
- 12 Dowsure-Fin-27B-V1 LLM 60.2
- 13
DeepSeek-V4-Pro Agent 52.7
Source: Official CNFinBench leaderboard
- 1
Doubao Seed 2.1 Pro LLM 83.7
- 2 Kimi K3 LLM 82.9
- 3
DeepSeek V4.1 Flash LLM 80.6
- 4 GPT-6.1 Sol LLM 80.6
- 5
Gemini 3.1 Pro LLM 80.0
- 6 Claude Opus 5.5 LLM 79.1
- 7 GPT-5.6 Sol Agent 78.1
- 8
Qwen 3.8 Max LLM 74.9
- 9
MiniMax-M3 LLM 69.5
- 10
DeepSeek-V4-Pro Agent 52.7
Source: Official CNFinBench leaderboard
| # | Entry | Score | Setting | Submitted by | Released | Run on | Source |
|---|---|---|---|---|---|---|---|
| 1 | | 83.7 | LLM | ByteDance | Jun 2026 | Oct 2026 | Official leaderboard |
| 2 | Kimi K3 | 82.9 | LLM | Moonshot | Jul 2026 | Oct 2026 | Official leaderboard |
| 3 | | 80.6 | LLM | DeepSeek AI | Sep 2026 | Oct 2026 | Official leaderboard |
| 4 | GPT-6.1 Sol | 80.6 | LLM | OpenAI | Sep 2026 | Oct 2026 | Official leaderboard |
| 5 | Intern-S2-Preview-397B open weights, 397B | 80.3 | LLM | Shanghai AI Laboratory | Oct 2026 | Official leaderboard | |
| 6 | | 80.0 | LLM | Feb 2026 | Oct 2026 | Official leaderboard | |
| 7 | Claude Opus 5.5 | 79.1 | LLM | Anthropic | Sep 2026 | Oct 2026 | Official leaderboard |
| 8 | GPT-5.6 Sol | 78.1 | Agent | OpenAI | Jul 2026 | Oct 2026 | Official leaderboard |
| 9 | MTFin-Qwen3.8-V2 closed weights, 27B | 76.0 | LLM | 美团金融大模型应用中心 | Nov 2026 | Official leaderboard | |
| 10 | | 74.9 | LLM | Alibaba | Aug 2026 | Oct 2026 | Official leaderboard |
| 11 | | 69.5 | LLM | MiniMax | Jun 2026 | Oct 2026 | Official leaderboard |
| 12 | Dowsure-Fin-27B-V1 closed weights, 27B | 60.2 | LLM | Dowsure | Nov 2026 | Official leaderboard | |
| 13 | | 52.7 | Agent | DeepSeek | Apr 2026 | Oct 2026 | Official leaderboard |
| # | Model | Score | Setting | Submitted by | Released | Run on | Source |
|---|---|---|---|---|---|---|---|
| 1 | | 83.7 | LLM | ByteDance | Jun 2026 | Oct 2026 | Official leaderboard |
| 2 | Kimi K3 | 82.9 | LLM | Moonshot | Jul 2026 | Oct 2026 | Official leaderboard |
| 3 | | 80.6 | LLM | DeepSeek AI | Sep 2026 | Oct 2026 | Official leaderboard |
| 4 | GPT-6.1 Sol | 80.6 | LLM | OpenAI | Sep 2026 | Oct 2026 | Official leaderboard |
| 5 | | 80.0 | LLM | Feb 2026 | Oct 2026 | Official leaderboard | |
| 6 | Claude Opus 5.5 | 79.1 | LLM | Anthropic | Sep 2026 | Oct 2026 | Official leaderboard |
| 7 | GPT-5.6 Sol | 78.1 | Agent | OpenAI | Jul 2026 | Oct 2026 | Official leaderboard |
| 8 | | 74.9 | LLM | Alibaba | Aug 2026 | Oct 2026 | Official leaderboard |
| 9 | | 69.5 | LLM | MiniMax | Jun 2026 | Oct 2026 | Official leaderboard |
| 10 | | 52.7 | Agent | DeepSeek | Apr 2026 | Oct 2026 | Official leaderboard |
Rankings by part
We read 21 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as 金融专业知识问答, Fin_Basics (Accuracy) and Fin_Cert_Exams (Accuracy). Doubao Seed 2.1 Pro leads 8 of them, and 5 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
金融专业知识问答
8 models we track have a score here.
- 1
Doubao Seed 2.1 Pro 92.7
- 2 Kimi K3 91.4
- 3
Gemini 3.1 Pro ↑2 90.3
- 4 Claude Opus 5.5 ↑2 89.7
- 5 GPT-6.1 Sol ↓1 89.0
- 6
Qwen 3.8 Max ↑2 88.0
- 7
DeepSeek V4.1 Flash ↓4 88.0
- 8
MiniMax-M3 ↑1 71.0
Fin_Basics (Accuracy)
8 models we track have a score here.
- 1
Doubao Seed 2.1 Pro 94.0
- 2 Kimi K3 92.0
- 3
Gemini 3.1 Pro ↑2 89.3
- 4 Claude Opus 5.5 ↑2 88.7
- 5
Qwen 3.8 Max ↑3 88.0
- 6 GPT-6.1 Sol ↓2 88.0
- 7
DeepSeek V4.1 Flash ↓4 86.0
- 8
MiniMax-M3 ↑1 76.0
Fin_Cert_Exams (Accuracy)
8 models we track have a score here.
- 1
Gemini 3.1 Pro ↑4 91.3
- 2
Doubao Seed 2.1 Pro ↓1 91.3
- 3 Kimi K3 ↓1 90.7
- 4 Claude Opus 5.5 ↑2 90.7
- 5
DeepSeek V4.1 Flash ↓2 90.0
- 6 GPT-6.1 Sol ↓2 90.0
- 7
Qwen 3.8 Max ↑1 88.0
- 8
MiniMax-M3 ↑1 66.0
金融业务理解与分析
8 models we track have a score here.
- 1
Gemini 3.1 Pro ↑4 80.9
- 2 Kimi K3 80.1
- 3 GPT-6.1 Sol ↑1 80.1
- 4
DeepSeek V4.1 Flash ↓1 80.0
- 5
MiniMax-M3 ↑4 78.7
- 6
Doubao Seed 2.1 Pro ↓5 74.9
- 7
Qwen 3.8 Max ↑1 71.4
- 8 Claude Opus 5.5 ↓2 66.5
Fin_Info_Extract (Micro-F1)
8 models we track have a score here.
- 1 Kimi K3 ↑1 62.3
- 2
Gemini 3.1 Pro ↑3 62.1
- 3 Claude Opus 5.5 ↑3 61.3
- 4
DeepSeek V4.1 Flash ↓1 61.0
- 5
Qwen 3.8 Max ↑3 60.7
- 6
Doubao Seed 2.1 Pro ↓5 60.5
- 7 GPT-6.1 Sol ↓3 60.1
- 8
MiniMax-M3 ↑1 54.5
Fin_Report_Parse (Accuracy)
8 models we track have a score here.
- 1 Kimi K3 ↑1 100.0
- 2
MiniMax-M3 ↑7 98.7
- 3 GPT-6.1 Sol ↑1 98.7
- 4
Gemini 3.1 Pro ↑1 98.0
- 5
Doubao Seed 2.1 Pro ↓4 98.0
- 6
DeepSeek V4.1 Flash ↓3 97.3
- 7
Qwen 3.8 Max ↑1 64.7
- 8 Claude Opus 5.5 ↓2 35.3
Fin_Reason_Consistency (Agent-as-a-Judge)
8 models we track have a score here.
- 1 Claude Opus 5.5 ↑5 72.4
- 2 GPT-6.1 Sol ↑2 71.4
- 3
Gemini 3.1 Pro ↑2 68.1
- 4
DeepSeek V4.1 Flash ↓1 66.1
- 5
MiniMax-M3 ↑4 64.9
- 6
Qwen 3.8 Max ↑2 62.5
- 7 Kimi K3 ↓5 59.3
- 8
Doubao Seed 2.1 Pro ↓7 41.9
Fin_Reason_Insight (Agent-as-a-Judge)
8 models we track have a score here.
- 1
Doubao Seed 2.1 Pro 99.0
- 2 Kimi K3 98.9
- 3
Qwen 3.8 Max ↑5 97.6
- 4 Claude Opus 5.5 ↑2 97.1
- 5
MiniMax-M3 ↑4 96.8
- 6
Gemini 3.1 Pro ↓1 95.5
- 7
DeepSeek V4.1 Flash ↓4 95.4
- 8 GPT-6.1 Sol ↓4 90.3
金融事实推理与计算
8 models we track have a score here.
- 1 GPT-6.1 Sol ↑3 83.3
- 2 Claude Opus 5.5 ↑4 80.5
- 3 Kimi K3 ↓1 79.3
- 4
Gemini 3.1 Pro ↑1 76.5
- 5
DeepSeek V4.1 Flash ↓2 74.8
- 6
Doubao Seed 2.1 Pro ↓5 74.0
- 7
MiniMax-M3 ↑2 69.0
- 8
Qwen 3.8 Max 56.6
Credit_Score (Accuracy)
8 models we track have a score here.
- 1 Kimi K3 ↑1 91.3
- 2 GPT-6.1 Sol ↑2 91.3
- 3
MiniMax-M3 ↑6 90.7
- 4
Qwen 3.8 Max ↑4 90.7
- 5
DeepSeek V4.1 Flash ↓2 90.7
- 6
Doubao Seed 2.1 Pro ↓5 89.3
- 7
Gemini 3.1 Pro ↓2 87.3
- 8 Claude Opus 5.5 ↓2 86.7
Quant_Invest (Accuracy)
8 models we track have a score here.
- 1 GPT-6.1 Sol ↑3 95.3
- 2
DeepSeek V4.1 Flash ↑1 92.0
- 3 Kimi K3 ↓1 91.3
- 4
Doubao Seed 2.1 Pro ↓3 88.0
- 5 Claude Opus 5.5 ↑1 88.0
- 6
Gemini 3.1 Pro ↓1 87.3
- 7
MiniMax-M3 ↑2 75.3
- 8
Qwen 3.8 Max 9.30
Loan_Analysis (Accuracy)
8 models we track have a score here.
- 1
DeepSeek V4.1 Flash ↑2 93.3
- 2 Claude Opus 5.5 ↑4 92.0
- 3 GPT-6.1 Sol ↑1 92.0
- 4
Gemini 3.1 Pro ↑1 90.0
- 5 Kimi K3 ↓3 89.3
- 6
Qwen 3.8 Max ↑2 88.0
- 7
MiniMax-M3 ↑2 82.7
- 8
Doubao Seed 2.1 Pro ↓7 79.3
BankAccount_Fraud_Detect (Accuracy)
8 models we track have a score here.
- 1 Claude Opus 5.5 ↑5 75.3
- 2 GPT-6.1 Sol ↑2 65.3
- 3
Doubao Seed 2.1 Pro ↓2 50.7
- 4 Kimi K3 ↓2 50.7
- 5
MiniMax-M3 ↑4 50.0
- 6
Qwen 3.8 Max ↑2 50.0
- 7
DeepSeek V4.1 Flash ↓4 50.0
- 8
Gemini 3.1 Pro ↓3 41.3
Fin_Num_Calc (Accuracy)
8 models we track have a score here.
- 1
Doubao Seed 2.1 Pro 59.1
- 2 Kimi K3 59.1
- 3
Qwen 3.8 Max ↑5 58.4
- 4 GPT-6.1 Sol 57.7
- 5
Gemini 3.1 Pro 55.0
- 6
MiniMax-M3 ↑3 54.4
- 7
DeepSeek V4.1 Flash ↓4 50.3
- 8 Claude Opus 5.5 ↓2 47.7
Fin_Reason_Align (Evidence-binding F1)
8 models we track have a score here.
- 1 GPT-6.1 Sol ↑3 98.0
- 2
Gemini 3.1 Pro ↑3 97.8
- 3 Kimi K3 ↓1 93.8
- 4 Claude Opus 5.5 ↑2 93.3
- 5
Doubao Seed 2.1 Pro ↓4 77.5
- 6
DeepSeek V4.1 Flash ↓3 72.4
- 7
MiniMax-M3 ↑2 61.1
- 8
Qwen 3.8 Max 43.1
金融合规与风险控制
8 models we track have a score here.
- 1 Kimi K3 ↑1 90.0
- 2
Doubao Seed 2.1 Pro ↓1 89.4
- 3 Claude Opus 5.5 ↑3 84.0
- 4 GPT-6.1 Sol 84.0
- 5
Gemini 3.1 Pro 83.7
- 6
DeepSeek V4.1 Flash ↓3 81.7
- 7
Qwen 3.8 Max ↑1 78.3
- 8
MiniMax-M3 ↑1 64.0
Fin_Compliance (Accuracy)
8 models we track have a score here.
- 1 Kimi K3 ↑1 91.3
- 2
Doubao Seed 2.1 Pro ↓1 90.0
- 3 Claude Opus 5.5 ↑3 86.0
- 4
Qwen 3.8 Max ↑4 85.3
- 5
DeepSeek V4.1 Flash ↓2 84.7
- 6 GPT-6.1 Sol ↓2 84.7
- 7
Gemini 3.1 Pro ↓2 84.0
- 8
MiniMax-M3 ↑1 64.0
Fin_Risk_Ctrl (Accuracy)
8 models we track have a score here.
- 1
Doubao Seed 2.1 Pro 88.7
- 2 Kimi K3 88.7
- 3
Gemini 3.1 Pro ↑2 83.3
- 4 GPT-6.1 Sol 83.3
- 5 Claude Opus 5.5 ↑1 82.0
- 6
DeepSeek V4.1 Flash ↓3 78.7
- 7
Qwen 3.8 Max ↑1 71.3
- 8
MiniMax-M3 ↑1 64.0
金融内生与应用安全
8 models we track have a score here.
- 1
Doubao Seed 2.1 Pro 87.7
- 2
Qwen 3.8 Max ↑6 80.1
- 3
DeepSeek V4.1 Flash 78.4
- 4 Claude Opus 5.5 ↑2 74.7
- 5 Kimi K3 ↓3 73.5
- 6
Gemini 3.1 Pro ↓1 68.5
- 7 GPT-6.1 Sol ↓3 66.4
- 8
MiniMax-M3 ↑1 65.0
Fin_Internal_Sec (Agent-as-a-Judge)
8 models we track have a score here.
- 1
Doubao Seed 2.1 Pro 87.1
- 2
Qwen 3.8 Max ↑6 79.4
- 3 Claude Opus 5.5 ↑3 77.1
- 4
DeepSeek V4.1 Flash ↓1 74.9
- 5 Kimi K3 ↓3 72.8
- 6 GPT-6.1 Sol ↓2 70.6
- 7
Gemini 3.1 Pro ↓2 70.0
- 8
MiniMax-M3 ↑1 67.3
Fin_App_Sec (Agent-as-a-Judge)
8 models we track have a score here.
- 1
Doubao Seed 2.1 Pro 88.2
- 2
DeepSeek V4.1 Flash ↑1 81.8
- 3
Qwen 3.8 Max ↑5 80.7
- 4 Kimi K3 ↓2 74.2
- 5 Claude Opus 5.5 ↑1 72.2
- 6
Gemini 3.1 Pro ↓1 67.0
- 7
MiniMax-M3 ↑2 62.6
- 8 GPT-6.1 Sol ↓4 62.1
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official CNFinBench leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| LLM | 83.7 | ||||
| Kimi K3 | LLM | 82.9 | |||
| LLM | 80.6 | ||||
| GPT-6.1 Sol | LLM | 80.6 | |||
| LLM | 80.0 | ||||
| Claude Opus 5.5 | LLM | 79.1 | |||
| GPT-5.6 Sol | Agent | 78.1 | |||
| LLM | 74.9 | ||||
| LLM | 69.5 | ||||
| Kimi K3 | Agent | 69.3 | |||
| Agent | 63.0 | ||||
| Agent | 60.2 | ||||
| Agent | 54.3 | ||||
| Agent | 52.7 |
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official CNFinBench leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| LLM | 83.7 | ||||
| Kimi K3 | LLM | 82.9 | |||
| LLM | 80.6 | ||||
| GPT-6.1 Sol | LLM | 80.6 | |||
| LLM | 80.0 | ||||
| Claude Opus 5.5 | LLM | 79.1 | |||
| GPT-5.6 Sol | Agent | 78.1 | |||
| LLM | 74.9 | ||||
| LLM | 69.5 | ||||
| Kimi K3 | Agent | 69.3 | |||
| Agent | 63.0 | ||||
| Agent | 60.2 | ||||
| Agent | 54.3 | ||||
| Agent | 52.7 |
2 models have held number one on the official leaderboard since its release in December 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official CNFinBench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official CNFinBench leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Law, finance and medicine
- MedXpertQA-MM In use Models answer expert-level medical exam questions that combine images with clinical records and test results. Scoring is accuracy.
Qwen 3.8 Max 80.4%
- HealthBench Professional In use 525 physician-authored tasks from real clinician chats (care consult, documentation, research) with physician-adjudicated rubrics. Claude Sonnet 5.5 69.2%
- HealthBench In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
- FinSearchComp In use Agents search the web to fetch live financial data, look up history, and run multi-step investigations, across 635 expert-written questions. Scored by answer accuracy.
Ling-3.1-flash 81.7%
- KorMedMCQA In use Models answer 7,469 multiple-choice questions from Korean licensing exams for doctors, nurses, pharmacists and dentists, held from 2012 to 2024. o1-preview 92.7%
- Harvey LAB-AA v1.1 In use Artificial Analysis's run of Harvey's Legal Agent Benchmark on Harvey's newer private set of 120 tasks across 24 practice areas. A deliverable with a material hallucination scores zero, and the headline is the share of judges who find every rubric criterion passed. GPT-6.2 Astra 15%
Data audit
All 9 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (13 entries).
- Pass Models scored 10 models we track, plus 3 other entries on its leaderboard.
- Pass Newest score The newest is from September 2026, for GPT-6.1 Sol.
- Note Frontier models Only 2 of the top 10 have a score. The strongest tested is Claude Opus 5.5 (#5).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on cnfinbench.opencompass.org.cn: 14 rows for models we track and 3 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.